Climate drift induced by the introduction of dynamic vegetation phenology

The addition of a new degree of freedom to a climate modeling system has the potential of inducing climate “drift” – while a modeling system may produce a stable and reasonable simulation of climate when some aspect of the system is fixed and specified realistically, allowing that aspect to evolve with the rest of the climate system may lead to significant changes in the model's climatology. Here we quantify the drift induced when a dynamic vegetation phenology model is coupled to an atmospheric general circulation model (AGCM). To produce the control climatology, the dynamic phenology model is first run globally offline, driven with observations-based meteorological forcing. The time series of vegetation state fields produced in this offline run are stored and then prescribed within a simulation of the full AGCM/vegetation model. That is, in the control simulation, the updated vegetation states produced by the phenology model are continually replaced by the states generated earlier in the offline simulation, so that the AGCM sees only “realistic” vegetation states, i.e., states consistent with realistic meteorology. To produce the “drift” climatology, the AGCM/vegetation model is run again, this time without the continual replacement of the vegetation states – the coupled system is allowed to go where it wants. Drifts in hydrological state appear to be relatively small, though those in the Sahel appear significant. Drifts in surface temperature, on the other hand, are larger and more widespread.